Dentification Algorithm Of Molten Pool Surface Characterization And Weld Penetration Status During Laser Deep Penetration Welding

Laser welding technology originated from the United States in1970. With the development of laser, laser power changed from small to large, quality and stability of laser beam improved, laser welding technology has been widely used in various industry fields. As a special welding process, the main advantage of the advanced laser welding is its large weld energy density in per unit length, high welding speed, small weld heat affected zone, small deformation, high mechanical properties of joints and weld profile dimensions do not subject to restrictions. We could get a large penetration depth used high power laser welding.Weld penetration is an important standard which evaluates the laser welding quality. However, the weld penetration can not be measured directly during laser welding. Controlling the weld penetration could only be carried out through an indirect estimation of weld penetration by other sensing signals. The research status and research methods of weld penetration status identification were introduced in this paper. Aiming at investigating the high weld-depth laser welding of type304austenitic stainless steel plate butt welding (the gap is less than0.1mm), an infrared sensitive high speed camera arranged off-axis orientation of laser beam was used to capture the dynamic thermal images of a molten pool. Paper tries to find the inner relation between molten pool surface characteristics and weld penetration status. lamge processing was used to extract the regions and characteristic parameters of keyhole and interested molten pool area. Aiming at the technical problem of molten pool region extraction, an Otsu threshold segmentation method which based on partily gray histogram was designed to solve it. The molten pool regions were corrected after threshold segmentation. The keyhole circularity, molten pool width, molten pool average gray and the variance of molten pool gray were defined as the characteristic parameters of molten pool, they also were normalized. Modeling method of fuzzy cultering were introduced, both Fuzzy C-Means cultering(FCM) and Gustafson-Kessel fuzzy cultering(GK) were used to build the model of weld penetration status identification. The effects of modeling methods and model parameters to the model identification rate were discussed. The FCM had a good identification rate of penetration class but not so well for none-penetration class. GK had a good identification rate of both classes. Finally, two effect parameter groups of GK were given. Paper focused on the relation of molten pool surface characteristics and weld penetration status, it provides some theoretical basis of weld penetration status identiflication real-time moniting during high weld-depth laser welding. It also could be used to weld quality real-time estimates after welding.